WO2001078451A1

WO2001078451A1 - Creating virtual surround using dipole and monopole pressure fields

Info

Publication number: WO2001078451A1
Application number: PCT/US2000/009556
Authority: WO
Inventors: John W. Norris; Mark Grzybek; Brad Wood
Original assignee: Harman International Industries, Incorporated
Priority date: 2000-04-10
Filing date: 2000-04-10
Publication date: 2001-10-18
Also published as: AU2000242253A1; CN1454446A; JP2004538669A; EP1275270A1

Abstract

The objective of virtual 3D sound technology is to convey to the listener an accurate impression of an acoustic environment. This is accomplished by conveying a realistic directional impression of sounds. Most of the successful techniques for 3D sound are based on the head related transfer function (HRTF). In this disclosure, we shall present a method that creates the desired impression using dipole and monopole pressure fields. The object is to recreate locally the pressure field an actual sound source would produce, in a neighbourhood of the listener's ears without using HRTFs.

Description

Creating Virtual Surround Using Dipole And Monopole Pressure Fields

BACKGROUND OF THE INVENTION

1. Field of Invention This invention relates generally to a method and system for creating virtual surround sound using dipole and monopole pressure fields using just the two front speakers for example and, more particularly, to recreate locally the pressure field an actual sound source, would produce locally without the sound source at that location.

2. Description of the Related Art With the introduction of multi-channel audio in DVD and on other media, there are situations where a listener would like to have the pleasure of listening to multi-channel audio, i.e., surround sound, without the inconvenience of having add rear speakers for the surround sound. For example, personal computers and televisions generally come with two speakers, to add surround sound effect, the listener generally needs to add rear speakers. In many instances this means that wires need to be run to the speakers, which add cost and sometimes can be esthetically unpleasing. Therefore, there is a need to create a sound pressure field that resembles the multi-speaker home theater system by using just the two speakers provided from electronic devices, such as from the personal computers and televisions. One way to create such a system is first to select the desired virtual loudspeaker positions. For example, the front loudspeakers could be at +/- 45 degrees, the center speaker would be at 0 degrees and the surrounds could be at +/- 90 degrees. Then the audio channel is filtered with the appropriate ipsilateral and contralateral positional filters, which can be derived from a head related transfer function (HRTF) measured impulse response. These filters give the listener the impression that the origin of the sound (i.e. the virtual loudspeaker) emanates from these positions in space. The system then filters each of the audio channels, sum all the left signals together and then sum the right signals to reduce them to a stereo pair. Because the system needs to cancel the cross talk associated with the left and right stereo loudspeakers, the system then filters the left and right signals with the inverse HRTF transformation associated with the real loudspeaker positions. To do so, inverse HRTF needs to be first calculated. However, one of the drawbacks of determining the inverse HRTF is that the pinna characteristics of the ears need to be measured, i.e., physical measurements need to be take, this can be a complicated process that takes time and money. Therefore, there still is a need for a method and system that creates the surround sound effect using two speakers in a computationally simple way without the need to calculate the inverse HRTF.

BRIEF SUMMARY OF THE INVENTION

A general feature of the present invention is to provide a method and system that creates the desired positional effect in a computationally simple way without the need to calculate the inverse head related transfer function (HRTF) in order to create surround sound effect using two speakers. An exemplary method and system according to the present invention creates the desired 3D sound impression using dipole and monopole pressure fields derived from simple point sources. That is, rather than using HRTF measurements for the surround sound effect, the present invention models the listener's ears as two points separated by a distance 2a, where a represents the listeners head radius. Thus, such effects as head diffraction may not be explicitly taken into account. One aspect of the present invention here is to recreate locally the pressure field an actual sound source would produce in a neighborhood of the listener's ears.

The exemplary method and system according to the present invention, eliminates one of the drawbacks of using HRTF, inter alia, which is using the, pinna characteristics of the ears to make the measurement. This means that the listener then "in effect" has to listen through these ears. Since pinna characteristics vary widely between individuals, the introduced notches and peaks associated with the measured pinnas may not correlate for a different listener. In contrast, with the present invention, the pressure fields near the listener's ears are approximated so that artificial peaks and notches in the transfer functions are no longer created, which colors the sound.

Another feature of the present invention is to steer the image around the listener to any desired position on a horizontal plane. This is accomplished by creating the necessary particle motion by decomposing the pressure field into its dipole and monopole parts.

Still another feature of the present invention is provide an efficient method of creating a sound pressure field, which resembles the multi-speaker home theater system through two speakers for Dolby ProLogic or Dolby Surround encoded materal.

In accordance with one aspect of the present invention, these and other features are accomplished by providing a method for creating surround sound using two front speakers, comprising the steps of: providing a sound source; decomposing the sound source into a dipole and a monopole terms; decomposing a pressure filed created by stereo speakers into the dipole and monopole terms; finding the inverse frequency responses to the dipole and monopole terms; calculating the approximation for the dipole according to

Δ =

providing a sum of the dipole and monopole signal to the left speaker; and providing a difference between the dipole and the monopole signal to the right speaker.

In accordance with another aspect of the present invention, these objectives are accomplished by providing a method for using dipole and monopole correction filters to virtualize Dobly ProLogic encoded stereo signals, comprising the steps of: providing a stereo ProLogic signal, wherein the stereo signal includes a left signal and a right signal; calculating the dipole by subtracting the right signal from the left signal; calculating the monopole by adding the left signal to the right signal; approximating the inverse monopole and dipole compensating transfer functions using low order filters; providing a signal to the left speaker that includes compensated dipole signal and the compensated monopole signal; and providing a signal to the right speaker that includes compensated monopole signal less the compensated dipole signal.

In accordance with yet another aspect of the present invention, these features are accomplished by providing a method for adjusting the gain (g) to adjust the width of the stereo enhancement, comprising the steps of: forming the monopole and dipole compensating filters for a stereo signal; and adjusting the gain of the compensating dipole filter to a desired value to vary the surround sound effect. In accordance with still another aspect of the present invention, these features are accomplished by providing a method for creating a surround sound using two front speakers with time delay, comprising the steps of: forming a first filter comprising: a filter FI, wherein the filter FI is a first transfer function from a speaker to the listener's closest ear; a filter LI, wherein the filter LI models the head shadowing of a listener's head from the speaker to the listener's furthest ear; forming a filter A comprising: a filter F2, wherein the filter F2 is a second transfer function from a desired source of sound position, wherein the sound source is described by a two by one column vector where: if the sound is to the right of the listener then the first value is the signal of the desired sound and the second column value is 0; if the sound is to the left of the listener then the first value is 0 and the second column value is the signal of the desired sound; a filter L2, wherein the filter L2 models the head shadowing of the desired source of sound position to the fiirherest ear of the listener; multiplying the column vector with H(z), wherein the

1 - z-^m'Z,, ^f 1 z~^m>L₂ ^Λ

#(_^•) = ^l ; approximating a time delay according to

\ -z-^2m'G(z) ^■ z" L 1 7~^m> I 1

(2aF smθ \ ml, wherein the m, = Round ; calculating the approximation for the dipole c J

\ — z^→HL according to Δ = — — - ; calculating the approximation for the monople according to

1 — z-'"'Z, l + z^~"'²L

Σ = — — - ; providing a sum of the dipole and monopole signal to the left speaker;

1 + z^~" ' Z, and providing a difference between the dipole and the monopole signal to the right speaker.

The above described and many other features and attendant advantages of the present invention will become apparent from a consideration of the following detailed description when considered in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

Detailed description of the preferred embodiment of the invention will be made with reference to the accompanying drawings.

Fig. 1 is an exemplary illustration of prior art using four speakers to create surround sound for a listener; Fig. 2 is an exemplary graph in accordance with one embodiment of the present invention fpr the function \A_Spk , assuming r₀ = l.Ow , = 0.09w , and θ = 20° ;

Fig. 3 is an exemplary graph in accordance with one embodiment of the present invention of monopole frequency response;

Fig..4 is an exemplary illustration in accordance with one embodiment of the present invention of a listener hearing a point source pressure;

Fig. 5 is an exemplary graph in accordance with one embodiment of the present invention of the plotted functions for |Δ _R | and |Σ_R |, representing dipole and monopole, respectively; Fig. 6 is an exemplary graph in accordance with one embodiment of the present invention representing the dipole and monopole (Δ and Σ) filters for speakers placed at +/- 10° and virtual source at 90°;

Fig. 7 is an exemplary graph in accordance with one embodiment of the present invention representing use of first order filters to approximate the low frequency behavior of the dipole and monopole (Δ and Σ);

Fig. 8 is an exemplary graph in accordance with one embodiment of the present invention representing use of filters to approximate the frequency behavior of the dipole and monopole (Δ and Σ) using 2 and 3 biquads; Fig. 9 is an exemplary graph in accordance with one embodiment of the present invention representing low frequency approximations for the dipole and monopole (Δ and Σ) filters for speakers placed at θ =-/+20, and a virtual speaker placed at φ=90;

Fig. 10 is an exemplary filter analog circuit diagram in accordance with one embodiment of the present invention; Fig. 11 is an exemplary illustration in accordance with one embodiment of the present invention of a stereo with a variable ground g;

Fig. 12 is an exemplary analog circuit in accordance with one embodiment of the present invention;

Fig. 13 is an exemplary filter block diagram in accordance with another embodiment of the present invention using two spealcers to create surround sound from a four channel sound card;

Fig. 14 is an exemplary illustration in accordance with the embodiment disclosed in Fig. 13, using two actual speakers to simulate surround sound from a four channel sound card; Fig. 15 is a exemplary filter block diagram in accordance with yet another embodiment disclosed in Fig. 13, using three speakers to support a four channel sound card;

Fig. 16 is an exemplary illustration in accordance with the embodiment disclosed in Fig. 15, using three actual speakers to simulate surround sound from a four channel sound card;

Fig.^' 17 is an exemplary illustration in accordance with still another embodiment of the present invention of using four front actual speakers to simulate surround sound from a four channel sound card; Fig. 18 is a exemplary filter block diagram in accordance with the embodiment disclosed in Fig. 17, using four front speakers to support a four channel sound card;

Fig. 19 is an exemplary graph illustrating a listener and associated transfer function; Fig. 20 is another exemplary graph illustrating a listener and associated transfer function;

Fig. 21 is yet another exemplary graph illustrating a listener and associated transfer function;

Fig. 22 is still another exemplary graph illustrating a listener and associated transfer function;

Fig. 23 is an further exemplary graph illustrating a listener and associated transfer function;

Fig. 24 is an another exemplary graph illustrating a listener and associated transfer function; Fig. 25 is an exemplary block diagram illustrating a listener and associated transfer function;

Fig. 26 is an exemplary analog diagram illustrating a listener and associated transfer function;

Fig. 27 is yet another exemplary block diagram illustrating a listener and associated transfer function;

Fig. 28 is still another diagram illustrating a listener and associated transfer function;

Fig. 29 is an exemplary graph of dipole and monopole cross talk filters;

Fig. 30 is an exemplary graph of concatenated dipole and monopole filters using time dealys;

Fig. 31 is another exemplary graph of spherical head model dipole and monopole filters;

Fig. 32 is another exemplary graph of a spherical head model;

Fig. 33 is an exemplary graph of a time delay model; and Fig. 34 is an exemplary graph of a point source model. DETAILED DESCRIPTION OF THE INVENTION

This description is not to be taken in a limiting sense, but is made merely for the purpose of illustrating the general principles of the invention. The section titles and overall organization of the present detailed description are for the purpose of convenience only and are not intended to limit the present invention.

Dipole and Monopole

As illustrated by way of example in Fig. 1, points Ptj and Pt₂ are spaced a distance

2a apart, where a is the radius of the listener's head 10 . The distances from Pti and Pt_∑ to the right front loudspeaker 14 are respectively r; and r₂. By geometry, this means that the left front speaker 12 and the right front speaker 14 are placed at a distance 2r₀ sin θ apart.

The surround loudspeakers, i.e., the left surround speaker 16 and the right surround speaker 18, cause particle motion to occur predominately in a direction perpendicular to the listener's ears, (assuming the listener is facing forward). Thus, if the two front speakers only can generate the same particle motion in a neighborhood of the listener's ears, the listener 10 would experience a similar sensation to that of the surround speakers playing this signal.

For example, if the left front speaker 12 is feed with a signal JFand the right front speaker 14 is feed with a signal -W, then these two speakers act as a dipole and cause particles to move backwards and forwards between them. In other words, the air motion is predominately perpendicular to the listener's ears. But since the dipole has a different frequency radiation characteristic than a normal speaker, the difference may need to be corrected. Once corrected, the listener should experience a convincing surround sound experience. The following is a discussion on how to correct for the frequency characteristic of a monopole and dipole.

A. DIPOLE AND MONOPOLE CROSS TALK CANCELLER FOR POINT SOURCES

All the front loudspeakers are treated as a point source, then the pressure at point Ptj (see Fig 1) due to the front loudspeakers may be described as:

similarly the pressure at the point Rt₂ may be described as:

where W=ve ; the wave number k = ωlc , ω is the angular frequency, cis the speed of sound, then:

^r\ - ψo + °^{2 +} 2^aro si ¹ θ > and r₂ = τjr₀ ² + a² - 2ar₀ sin θ .

Moreover, the magnitude of the pressure may be described as: |Δ_S .| =|R(Pt,)| = |R(pt₂)| .

As illustrated by way of example, the function Δ₅ . may be plotted as shown in Fig. 2, assuming r₀ =\.o_m, = 0.09m , and θ = 20° . (See Fig 1 for definition of variables.)

Assuming that _r„ » a , i.e-_> that the distance between the speaker and the listener is much greater than the radius of the listeners head, then using the binomial expansion one ordinarily skilled in the art can derive that:

where p = ^a¹ + r₀ ² . Therefore:

(1-2)

So sound with wavelength λ that satisfies

λm = as θ . where m is an integer, give rise to notches in the frequency domain at the frequencies:

mc

Λ = a nθ

Thus, there is a need to compensate for these notches and for the low frequency cancellation of the dipole. To compensate for this behavior, the transfer function is inverted. This means that pre-filtering the signal with the invert filter particle motion at Pti and Pt₂ in Fig. 1, results in a very similar particle motion that is created by the surround loudspeakers. This means that the frequency response is also correctly compensated for.

By feeding the same signals ue^,ωt into the left and right front loudspeakers, the magnitude of the pressure at the points Ptj and Pt₂ may be expressed as:

Using (1.1), the above equation (1.3) can be further expressed as:

(akr.

^JSpk (*)| = 2 cos ±-smθ l + O (1.4) p²

Fig. 3, illustrates by way of example plotted the monopole frequency response.

Here th liee nnoottcchheess aappppeeaarr ffoorr tthhee wwaavveelleennggtthhss tthhaatt ssaattiissffyy

(2m + iμ = 4αsi 0 ,

Thus, the notches appear at the frequencies

(2m + l)c

JM — ^' 4αsin#

Here, the cross talk of the front loudspeakers for a monophonic signal can be corrected by inverting this transfer function and using it to filter the input signal u. In other words, the cross talk associated with the front loudspeakers playing monophonic and out of phase signals (dipole) can now know be corrected. Moreover, the dipole terms cause particle motion that is substantially perpendicular to the listener's ears and the monopole term generates particle motion that is tangential to the listener's ears. By weighting these components, the method and system according to the present invention can steer sounds to any desired position.

B. POSITIONING A SOUND SOURCE

As illustrated by way of example in Fig. 4, when the listener hears a point source xe¹ , the pressure at that point Pti may be described as:

-iklϊ.

P(Pt, ) = xe"

similarly the pressure at Pt₂ may be described as:

M

P(Pt₂) = xe"

R> By subtracting and adding the above pressures, they may be described as the following:

and

Fig. 5, illustrates by way of example, the plotted functions of |Δ _R ] and |Σ_R |

Assuming that R₀ » , i.e., that distance between the speaker and the listener is much greater than the radius of the listeners head, then using the binomial expansion:

and

where p = τja² + R₀ ² . Therefore

By decomposing the sound source into a dipole and a monopole terms, in accordance with the present invention, the ratio of the magnitudes of dipole and monopole may be described as proportional to the direction of the sound. As discussed above, the positions of the notches are related to the direction of the sound. It is conceivable that the brain can add and subtract these signals to detect the variations in the comb frequency behavior above. A listener noticing the frequency location of these notches can determined, the direction of the origin of the sound.

Using two speaker to give the impression that the sound is emanating from the direction φ, the system according to the present invention compensates for the cross talk associated with the dipole and monopole from the two speakers. To do so, the system divides the dipole term by [Δs_pk J and the monopole term by [∑s_Pk] to obtain

and

∑ = . ^{∑ R} (0. 11)

Fig. 6, illustrates by way of example plots representing the dipole and monopole (Δ and Σ) filters for loudspeakers placed at +/- 10° and virtual source at 90°. Therefore, the method and system according to the present invention creates the effect that the sound is emanating from the angle φ using the stereo loudspeakers at ± θ , by feeding the left front speaker with the signal

£_$* = ∑+ Δ (0. 12) and the right front speaker with the signal:

R_Sp* = ∑- Δ . (0. 13)

The sound the listeners hears, assuming that the speakers behave as point sources, is:

Then, substituting in the expansions for rj and r₂ the equation may be further described as:

l~ ^J*Eva„r~ — (0. 15)

and

R Ear - (0. 16)

Rearranging equations (1.12) and ( 1.13), the following may be derived: = L_S≠ + RSpk and

Δ - L_Spk R_Spk , so equations (1.15) and (1.16) can be written as:

(0. 17)

and

kp

R_Eaι. = (∑ cos(ka sin θ) + z^'Δ sm(ak sin #)) + O (0. 18)

P p² Using the asymptotic expressions for ∑s_pι_< , Δs_Pk, ∑R &nd AR (equations (1.2), (1.4), (1.7) and (1.8)), the following may be derived:

and sin(α&sin^) limit at p-+0 ΔU (0. 20) sin(aksmθ)

Assuming that Rø=rø and ro>>α. Substituting in these expressions into the equations (1.17) and (1.18), the following may be derived:

L_Ear =

and

Which we can express as:

and

From the above one ordinarily skilled in the art can derive that there is a delay of:

τ = 2 αsin (0. 23)

between the signal heard at the left and right ear. From Rayleigh [4], such a delay is known to give directional cues. Thus, the low frequency approximations to Σ and Δ to the signals to the left and right ears may be delayed by the desired amount to give the listener the impression that the source is at an angle <j> using loudspeakers at +/- θ.

There are several ways to implement a delay to create the desired pressure fields at the listener's ears to give the impression of surround sound. In accordance with one aspect of the present invention, a first-order shelving filter may be used to approximate Σ and Δ, so that there is no need for extra memory to implement a delay. In accordance with another aspect of the present invention, with the filters, they may be implemented as simple analog filter circuits rather than using analog components to implement the the delay, which is generally more expensive to implement than the filters.

C. SIMPLE APPROXIMATION TO POINT SOURCES

In one embodiment of the present invention, a low frequency approximation to Δ and Σ is implemented by using equations (1.3),(1.4), (1.5) and (1.6) from above.

and

That is, using equations (1.5), (1.6), (1.7) and (1.8), from the above one ordinarily skilled in the art can derive that the asymptotic behavior of Σ and Δ may be represented as:

Limit k→O |∑| = ι (0.1) a t p~>0 and sin$zi

Limit k→O (0.2) θ/ P→0 H sin#

In other world, in accordance with the present invention, it is demonstrated that the low frequency behavior of Σ and Δ is independent of the listener's head size. In addition, both Σ and Δ are flat for low frequencies; as illustrated by way of example in Fig. 7. That is, in this embodiment, the present invention uses simple first order filters to approximate the low frequency behavior of Δ and Σ. Fig. 7 illustrates by way of example the plotted simple approximations of both Δ and Σ (including the case where Σ is approximated by a constant).

In accordance with the present invention, a good image is obtained by approximating just the low frequency behavior of Δ and Σ. In other words, by approximating just the low frequency end, a good approximate image may be obtained of actual frequency behavior of Δ and Σ. This allows a first order shelving filter to be used and give a good approximation for |Δ| up to 1.5 kHz. As Σ is flat for frequencies up to approximately 400 Hz, so that this can be approximated by a constant. Alternatively, a second order notch filter may give a better approximation, as illustrated in Fig. 7. Fig. 8 illustrates by way of example, a filter approximation for Σ and Δ which uses 2 and 3 biquads. To obtain a better approximation more biquads may be used. In accordance with another embodiment of the present invention, high frequency approximation behavior may not be needed, as the point source model becomes inaccurate. However, it is within the scope of the present invention to do so if desired. Note that in accordance with the present invention, the smoothed point source model substantially gives a good approximation to Rayleigh's spherical head model for frequencies up to about 10 kHz. See, Morse P., Ingard K. 1986 entitled "Theoretical Acoustic" Princeton Press, for explanation of Rayleigh spherical head model.

As discussed above, by implementing the approximations to Δ and Σ, the method according to the present invention generates the required delay between the left and right ear to generate the surround sound using two speakers. Moreover, since the Δ and Σ can be represented as simple filters, the method according to the present invention can build analog circuits to give this spatial audio effect, rather than implementing a delay in an analog circuit that is generally more complex. In other words, the present invention uses the filters to represent the approximation of the Δ and Σ. One of the advantages of using the approximation of the Δ and Σ is that the approximation is easier to calculate and implement using filters, rather than using analog circuits to represent the actual Δ and Σ.

Moreover, with the method and system according to the present invention, the apparent location of the sound may be varied by adjusting the DC height of the low-pass shelving filter as determined by equation (2.2). For example, to steer the sound to 90 degrees using loudspeakers placed at +/- 20 degrees, the DC offset needs to be sin (90)/sin (20)=2.92, i.e., nearly three times the value of Σ. Fig. 9 illustrated by way of example plots of Δ and Σ for speakers, i.e., point sources, placed at +/-20 degrees and a virtual loudspeaker, i.e., a point source, at 90 degrees and the above simple approximations, to illustrate the correlation.

Alternatively, a shelving filter calculated for 90 degrees may be used and real speakers placed at the angles +/- θ. Then the shelving filter may be used to steer the sound

to another angle, say φ by adjusting the DC offset of the dipole filter Δ to — — . The sin# knee of the shelving filter is generally a function of the angle φ but the low frequency positioning is generally determined by the DC offset of the dipole filter Δ(0). With this in mind, the present invention creates a simple analog circuits that can be used to steer the position of the sound by adjusting a resistor value. Fig. 10 illustrates by way of example, an analog circuit that creates this effect Moreover, filters may be implemented in the filter block diagram as illustrated by way of example in Fig 10.

As illustrated by way of example in Fig 10, the filters in blocks Δ and Σ may be first order shelving filters. If desired, a listener can also adjust the parameter g. By doing so, i.e., adjusting g, the impression of auditory envelopment may be varied, this affects the DC offset of the dipole filter (Δ(ω)) compared to the monopole, thereby effectively steering the apparent source of the sound to the angle φ=arcsin(g*Δ(0)*sin θ); as illustrated by way of example in Fig. 11. In this embodiment, the speakers may be assumed to be at +/- θ with respect to the listener.

In an another embodiment of the present invention, the filter block diagram may be implemented digitally or as a simple analog circuit using 4 op-amps, as illustrated by way of example in Fig. 12. (WHERE IS FIG. 12)

Yet another embodiment of the present invention is to support four-channel sound cards for personal computers (PCs). Although four channel sounds are becoming popular there are times when it would be inconvenient for the listener to have four loudspeakers surrounding him. In other words, the listener has to purchase two more speakers and would have to run the wires to the speaker. Moreover, with lap top computers that are portable, having surround speakers behind the listener in many instances is not a viable option.

Accordingly, as illustrated in Fig 13, a filter block diagram is shown that may be used to create the illusion of four loudspeakers surrounding the listener using just two loudspeakers. As illustrated by way of example in Fig 14, two actual speakers can simulate a surround sound effect as if the surround sound is coming from four virtual speakers.

Still another embodiment of the present invention is to create surround sound effect from three speakers. As illustrated by way of example in Fig 15, another filter block diagram is shown that may be used to create the illusion of four speakers surrounding the listener, yet with only three speakers. This is illustrated by way of example see Fig 16, here along with the two front speakers, a third speaker is placed behind the listener to give a strong rear image. In particular, the third loudspeaker may be coupled to the listener's chair substantially behind the listener's head.

Even further, another embodiment of the present invention is to simulate surround sound using four speakers in front of the listener supporting a four channel sound card, as illustrated by way of example in Fig. 17. To do so, Fig. 18 illustrates by way of example yet another filter diagram that may be used to create the surround sound from four speakers in front of the listener using four channel sound card.

With regard to implementing the block diagrams, they may be implemented through a variety of methods that is known to one ordinarily skilled in the art. For example, the block diagrams may be implemented through analog circuits as well as digital. In the digital domain, the source of the sound may be steered to any desired location using the first order filters, for example.

Still another feature of the present invention is provide an efficient method of creating a sound pressure field, which resembles the multi-speaker home theater system through two speakers for Dolby ProLogic or Dolby Surround encoded material. For example, by plotting |P(Ptι)| as a function of frequency, the low frequencies may be suppressed, as illustrated by way of example in Fig. 19. (This curve is labeled dipole in Fig. 4.) To compensate for this radiation frequency behavior, the low frequencies of the signals v and -v are boosted. If this is done correctly, the present invention creates a flat response for the listener, giving the impression of surround loudspeakers while still retaining the desired particle motion. To correctly compensate for this cancellation, the present invention inverts the expression for |P(Pt])|; see Fig. 20. As illustrated by way of example in Fig. 20, a number of different low frequency boost corrections for four different loudspeaker angles have been plotted. That is, as the real loudspeaker's angle θ decreases (that is, they are moved closer together) the low frequencies' boost ( ' _Boo_st) may need to be increased, as one ordinarily skilled in the art would expect.

To find an approximation for the pressure at the listener's ears, the following

assumptions may be made: that a « r₀ and define p = r — + a² . The pressure at the

listener's left ear may be found to be:

P ( Pt , )

By expanding ri and r₂ as a binomial expansion in terms of — the required low frequency

P boost may be expressed as: ψ

S b,oon 2sin( sin#)

If the signal L_Spk = S_BoostV and R_Spk = -Sβ_0osiv through the left and right loudspeakers

ear L Spk ear R Spk J

respectively, then the listener hears assuming that the loudspeakers act as point sources. a

By again expanding ri and r₂ in terms of — P

e^ikp a²

L = ((L spk , + R ^{" '}spk , ) ' cos(ka sw θ) - i(L spk , ~ R spk ' sm(k sin θ)) + 0(-_τ) (LA and

e'^kp a²

Since £§,* and Rs_pk are out of phase for a dipole,

and

ipy _ r _ D sιn(λαsιn0)

Substituting these expressions into equations ( A) and (2.A), then

7 = p^ikP_V and

Rear = ~e^ikPV

These are the signals the listener would hear if the signals v and -v were played out of surround loudspeakers placed a distance p away from the listener.

By boosting the low frequencies and playing the out of phase signal into the right and left front loudspeaker then the listeners should have a similar acoustic impression to that of surround loudspeakers placed at a distance p with respect to the listener playing the same signal.

In other words, the dipole term creates a pressure field that corresponds to the surround pressure field once the low frequencies have been boosted. The particle motion in this case is predominately perpendicular to the listener's ears. To create a component tangent to the listener's ears the method and system according to the present invention plays the same signal from the right and left front loudspeakers, i.e., u e^ml. That is, the pressure at Pti and Pt due to u e'^ωt is:

Fig. 19 illustrates, by way of example, the magnitude of this function as a function of frequency for loudspeakers at +5°. (This curve is labeled "monopole" in Fig. 4). From the above, one ordinarily skilled in the art can derive that the asymptotic behavior of the correction transfer function for the monopole signal is:

2cos(fc.cos#)

Fig. 21 illustrates, by way of example, the plotted monopole and dipole correction filters for loudspeakers placed at ±5°, ±10° and ±15°. In other words, as the loudspeakers are moved closer together, the required boost is increased.

To create the correct surround and monopole terms, the method and system according to the present invention feeds the left and right loudspeakers the signals

Ls_Pk = Mu + Dv and Rspk ^ Mu -Dv where

= - P (4.A)

2cos(ka nθ)

and

D = ^ (5.A)

2sin(/Vαsm#)

From the above, one ordinarily skilled in the art can express the equations (l.A) and (2.A) as:

e^ilcp ( -

L = (M u cos(ka sin θ) - iDv sin(ka sin θ) + O ~ I (6.A)

and (7.A)

Therefore, the listener hears

and

In other words, the sound pressure field at the listener's ears has been decomposed into two different components, a dipole and a monopole. By adjusting the relative magnitudes of the monopole with respect to the dipole, the method and system in accordance with the present invention affects the impression of the surround effect. For example: (10.A)

M = « -

^ 2 cost Arα sin 0)

then by making g greater than unity, the monopole would dominate over the surround, giving the impression of a reduced sound stage. Conversely, by making g less than unity, then the spatial sound stage is increased. Thus, by varying g, the method and system in accordance with the present invention can adjust the surround effect.

Fig. 22 illustrates, by way of example, the plot of the correction filters for the monopole signal for loudspeakers placed at ±5°, ±10° and ±15°. Since these correction filters are relatively simple approximation may be used over a significant portion of their frequency range by simple first order shelving filters. Fig. 23 illustrates, by way of example, plotting of the correction filters for ±5° and the simple first order approximations to them. The approximation is a filter L(z) of the form

Since the correction filter for the monopole signal is generally flat over much of its frequency range, it may be multiplied by a constant to simplify the required filtering further. As illustrated, by way of example, in Fig. 24, another approximation for a ±5° correction filter may be given when the monopole term is taken to be a constant and the approximation to the dipole correction filter substantially approximates the dipole filter above 200 Hz. One of the benefits of this approximation is that the poles of the first order shelf filter may be moved inwards away from the unit circle for the dipole correction filter. This helps reduce the rounding error if the filter is implemented in fixed-point arithmetic. D. SIMPLE VIRTUALIZATION OF DOLBY SURROUND AND STEREO

ENHANCEMENT

The Dolby Surround or ProLogic encoding process comprise mixing the multichannel data into a stereo pair

and

where the total left channel L_t includes the surround, left front and center channels and the total right channel R_t includes surround, right front and center. Note that Dolby recommends that the stereo down mix of the 3/2 Dolby Digital (AC3) mode should be and

where C, R^, _F, R_S, and is are respectively the center, right front, left front, right surround and left surround channels.

To provide a surround sound experience, the present invention makes the surround channels seem to come from the listener's sides and the front channels to come from the front. That is, the surround signal to seem to come from the sides, as the method and system according to the present invention, boosts the low frequencies of the diffe ence^ signal as described in the last section. To do this, the sum and difference of the Dolby Surround (ProLogic) is formed. The difference generates the dipole term and the sum the direct or monophonic component. By subtracting R_t from L_t the following dipole expression is obtained:

v = Lt - Rt = {L - R) - 2$;

and by summing L_t and R_t, the following monopole expression is obtained:

The low frequencies of the L_f-R_t signal may be boosted to compensate for the radiation characteristics of the dipole by filtering this signal with the low frequency shelving filter LPD- AS illustrated by way of example in Fig. 24, the listener should experience the surround channel enveloping him in the same way a properly decoded Dolby Surround channel would if played back using surround spealcers. Any sounds that come from the difference between the front left and right channels may also be boosted at the low frequencies. To correct the monopole signal, the present invention filters it using the high pass shelving filter HPM as shown by way of example in Fig. 24. The signals fed to the left and right loudspeakers may be expressed as

and

Rspk — HPM ^U - LPD ^V< (15.A)

Since HPM and LPQ approximate

p v v =

2 ύu(ka ύn $)

and p u

M =

2 cos(7_*α sin 0) ^'

then the equations (6. A) and (7. A) may be used to show the effect of these transfer functions on the front channels. From the above, one ordinarily skilled in the art may derive that:

ea = C^if>{L + V5C - 25) + 0 _, (^16A>

and

n_car

+ 2 ) + o ( j . °^7A)

That is, the present invention removes the cross talk associated with the left and right front channels that the listener would hear if a multi-loudspeaker system were used to play back the decoded Dolby Surround or ProLogic signals. This may be implemented using the filters LP_D and HP_M digitally or as a simple analog circuit. Fig. 25 illustrates for example a filter block diagram representing the above. Moreover, one embodiment of a possible analog circuit implementation for this filter structure is illustrated in Fig. 26.

If the above filter structure is applied to a normal stereo signal, then any out of phase signal between the left and right channels is substantially boosted and creates the impression of envelopment. Alternatively, a part of the original left and right channel are mixed and may be, as illustrated for example in Fig.27, back into the output. Fig.28 illustrates, by way of example, the listener and the associated transfer function. E. USING TIME DELAYS

In the above sections, the present invention models the loudspeakers and monaural sounds as point sources. To compensate for the cancellation of sound at certain wavelengths (frequencies), another model that may be used is time delays to find expressions for the contralateral and ipsilateral transfer functions. Consider Fig 1 again, and assume that the contralateral and ipsilateral transfer functions may be expressed by:

A - FLz-

and

S = F

The integer mi equals:

(TaF. smθ^' m_x - Round]

where c is the speed of sound, and Fs is the sample rate. Lj is a low pass shelving filter that models the shadowing effect of the head. F models the ipsilateral transfer functions. Therefore:

The inverse to this matrix is: ' 1 ~ z-"'< L^

F{l -z-^2m'L,²) - -T""Z, 1

Since F affects both signals equally, it may be set to unity. The cross talk cancelled signal is therefore given by:

If the low pass shadowing effect is not included to the head, the

poles on the unit circle may be at e^**¹"'¹ . Thus, the low pass filter L(z) stabilises the cross talk canceller and models the high frequency attenuation of sound. This low pass filter is designed to move the poles inwards away from the unit circle. In the following letting:

_{m =} ^bo + *l

1 + a,z^'

By summing and subtracting the signals X and X_R then:

1

Δ~ SpkT = ^■

\ -L,(z)z- and

1

∑~ SpkT = ^■ l + L, (z)z^~" Fig. 29, illustrates the plotted function Σ^-1s_P τ and Δ^-1s_Pkτ •

Again assuming that a sound x is at an angle φ with respect to the listener, the ipsilateral and contralateral transfer functions are given by:

A_φ = FL₂z^' and

S_Φ = F . Hence the cross talk cancelled positional filters may be expressed as: -z "'>f^λ - i-₂

H(z) = -

"G(z) l^-z-""L, 1 J ¹ J

Therefore, the positional dipole and monopole filters are

X - z^'^

Δ =

^>L, and

l + z-" L ' where G(z)=L](z)²... Fig. 30, illustrates plotting of Σ and Δ for a virtual source at 90 degrees and loudspeakers at +/- 20 degrees.

F. SPHERICAL HEAD MODEL

To see how good these approximations are, one can compare them to Rayleigh' s [5] spherical head model. Using a spherical head model to calculate the transfer functions from loudspeakers at ±20° and a sound at 90° to positions at ±120° on the sphere (to represent the positions of the ears) as illustrated in Fig. 31. That is, both the time delay (Fig 30) and the point sources give reasonable approximations to these filters (Fig 31) up to 2 kHz. Note that these approximations continue to improve as the actual loudspeakers are moved closer together. For example, if the actual loudspeakers are placed at ±io' then the point source and time delay approximations are reasonable up to about 6 kHz, see Figs 32, 33 and 34.

By measuring HRTFs for a particular directions and calculate the concatenated sigma and delta filters then the resulting filters give positional information but tend to make the resulting virtual image sound hollow and coloured. To improve these filters, Cooper Bauck [Cooper D., Bauck 1989. Prospect for transaural recording. J. Audio. Eng. Soc. Vol. 37, pp. 3-19.] devised an equalisation method. To implement a Cooper Bauck equalisation, the monopole and dipole filters must be divided by the absolute magnitude of the combined filters. Using the Cooper Bauck equalisation to calculate the cross talk cancelled and positional filters, then: Δ_κ E_Spk

and

where E = |Δ|² + 1∑|² . In other words, if both Sigma and Delta have peaks or notches in the same region then this equalisation will flatten out these undulations. In particular, the spatial cues associated with the localising bands will cause both Sigma and Delta to be reduced (or increased) in magnitude in certain frequency bands. Therefore this equalisation may destroy some of the spatial information that may help resolve some of the ambiguity associated with the cone of confusion. Thus, it removes most of the spectral notches associated with the pinna.

The point source model and the time delay models give reasonably good approximations to the spherical head model of concatenated positional and cross talk cancelling filters.

Although the present invention has been described in terms of the preferred embodiment above, numerous modifications and/or additions to the above-described preferred embodiments would be readily apparent to one skilled in the art.

In closing, it is noted that specific illustrative embodiments of the invention have been disclosed hereinabove. However, it is to be understood that the invention is not limited to these specific embodiments. With respect to the claims, it is applicant's intention that the claims not be interpreted in accordance with the sixth paragraph of 35 U.S.C. § 112 unless the term "means" is used followed by a functional statement.

Claims

What is claimed is:

1. A method for creating surround sound using two front speakers, comprising the steps of: providing a sound source; decomposing the sound source into a dipole and a monopole terms; decomposing a pressure filed created by stereo speakers into the dipole and monopole terms; finding the inverse frequency responses to the dipole and monopole terms; calculating the approximation for the dipole according to Δ = — — ;

calculating the approximation for the monopole according to Σ = — — ;

2. A method according to 1, wherein the sound source is form a predetermined location that is at distance R from the listener.

3. A method according to 1, wherein the decomposing is done for every location of the sound source.

4. A method according to 1 , wherein the approximation is done by a filter.

!

5. A method according to 4, wherein the filer is a low order filter.

6. A method for using dipole and monopole correction filters to virtualize Dobly ProLogic encoded stereo signals, comprising the steps of: providing a stereo ProLogic signal, wherein the stereo signal includes a left signal and a right signal; calculating the dipole by subtracting the right signal from the left signal; calculating the monopole by adding the left signal to the right signal; approximating the inverse monopole and dipole compensating transfer functions using low order filters; providing a signal to the left speaker that includes compensated dipole signal and the compensated monopole signal; and providing a signal to the right speaker that includes compensated monopole signal less the compensated dipole signal.

7. A method according to 6, wherein the dipole is a first order filter.

8. A method according to 6, wherein the monopole and the dipole filters do not necessarily tend to the same value at high frequency.

9. A method according to 6, wherein the filters are calculated using point source approximation.

10. A method according to 6, wherein the dipole compensating filter has a low frequency boost greater than 10 db.

11. A method according to 6, wherein the stereo signal is a ProLogic signal.

12. A method for adjusting the gain (g) to adjust the width of the stereo enhancement, comprising the steps of: forming the monopole and dipole compensating filters for a stereo signal; and adjusting the gain of the compensating dipole filter to a desired value to vary the surround sound effect.

13.' A method according to 12, wherein the gain is adjustable between 0.0 db and -20.0 db.

14. A method according to 12, wherein the stereo signal is a ProLogic signal.

15. A method for creating a surround sound using two front speakers with time delay, comprising the steps of: forming a first filter comprising: a filter FI, wherein the filter FI is a first transfer function from a speaker to the listener's closest ear; a filter LI, wherein the filter LI models the head shadowing of a listener's head from the speaker to the listener's furthest ear; forming a filter A comprising: a filter F2, wherein the filter F2 is a second transfer function from a desired source of sound position to the listener's nearest ear, wherein the sound source is described by a two by one column vector where: if the sound is to the right of the listener then the first value is the signal of the desired sound and the second column value is 0; if the sound is to the left of the listener then the first value is 0 and the second column value is the signal of the desired sound; a filter L2, wherein the filter L2 models the head shadowing of the desired source of sound position to the furthest ear of the listener; multiplying the column vector with H(z), wherein the v z-"^hL,

H(z) =

\-z-^G(z)\ -z ,₂- 1 approximating a time delay according to ml, wherein the

(2aF_i sinfl^' m, = Round

16. A method according to 15, wherein the compensating filters may be

expressed as monopole and dipole filters as given by Δ = — — - and Σ =

respectively.